Early life reliability growth testing with non-constant failure intensity

Abstract

Early-life reliability is a key characteristic for complex technical products as it addresses the very first time-span of product use. Any unreliability in this phase jeopardizes both customer satisfaction, as well as warranty costs and finally the economic success. Review of several early-life warranty cases shows a mix of classical reliability issues, i.e., problems occurring during usage time, and quality related issues that often manifest themselves during initial inspections by product quality or even the customer, i.e., at usage time t = 0. Classical reliability growth models do not explicitly consider these two different sources of unreliability but require zero failures at time zero and provide time-dependent reliability estimates. We present a model which shares the concept of time-dependent reliability measures but considers in addition the initial quality issues. By taking advantage of well-known reliability growth models and basic statistical principles we were able to provide a flexible approach useful for practical application.